Paper Eliciting Expectations

We document how variation in question format used to elicit beliefs of percentage changes, e.g. return expectations, influences reported beliefs. We provide a ground truth distribution and vary whether participants directly report beliefs or build a distribution. Participants accurately report differences in means and volatilities irrespective of method. For levels, directly asking for means induces large positive bias and significant noise. Volatilities from confidence intervals are unreliable. Reporting full distributions yields more accurate means and volatiles. Our analysis highlights the validity and limitations of survey evidence for finance applications and helps researchers effectively design surveys and interpret their results.

Read the Working Paper